Bandwidth Based Design Methodology for Wiener Filters for Online Signal Denoising

Online signal denoising refers to a digital filtering techniques in order to estimate a signal s[n] from the noisy signal x[n] = s[n] + v[n], where v[n] is noise. Standard filtering techniques lead to a time delay of the signal estimate ŝ[n], which is caused by the group delay of the filter. Wiener filtering is one possible way to reduce this delay. The design of a Wiener filter requires a statistical model of the signal s[n] and the measurement noise v[n]. The noise model can be determined by means of signal analysis of the measurement noise v[n]. The modeling of the signal s[n] on the other hand requires more dedicated analysis. This makes the application of this filter less practicable. In this paper a design methodology for Wiener filters is presented. The modeling of the signal is done by means of the signal bandwidth, hence the design effort is equal to the design of a standard digital filter. Also the same design tools, e.g. a filter design methods, can be used. The performance of the approach is presented by means of simulation studies and a measurement example.